Main site

Operations AI agent · Travel and expense

Travel and Expense Audit AI Agent

Test a proposed trip against the duties that land before departure, name the human each destination wants, and hold the claim arriving weeks later against the trip that was declared.

Typical delivery
4–6 weeks
Deployment
Your stack
Mobility lead
Before the trip
After launch
Agent Care

What this agent does

Prepares the trip, never declares it

In

  1. 01

    A trip is proposed, and the destinations, the dates and the work to be done are read before booking.

  2. 02

    A country is entered, and Regulation 883/2004 sets a ceiling with no floor, so a one-day meeting counts.

Reason

  1. 03

    A visit is called a business trip, and that carve-out sits in an April 2026 agreement not yet law.

  2. 04

    A service is delivered rather than received, and the carve-out being drafted would not reach it anyway.

  3. 05

    An A1 is held, and it binds solely in the area of social security, so the declaration is separate.

Decide

  1. 06

    A deadline runs off national law, and Switzerland forbids starting work for eight days after notice.

  2. 07

    A destination wants a person present, and that name is captured in the declaration itself.

Out

  1. 08

    A claim arrives weeks later, lawfully, and it is read against the trip declared, not instead of it.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

Reading the duties, dating the clocks and holding the record belong to the agent. Filing and clearing belong to a named mobility lead, who decides.

Example workflow

One trip, proposal to clearance

  1. 1Trip proposedBooking request, calendar invitation, project record or a travel-management feed
  2. 2Duties dated from the destinationThe countries, the work to be done there, the earliest deadline each one sets and the lead who holds it
  3. 3Evidence assembledThe A1 request, the declaration draft, the person each country wants named and the dates
  4. 4Controls appliedDestination checks, deadline checks, named-person checks and completeness confidence

No human action required

Stages 1 to 4 run unaided, and nothing is filed at any of them — the agent is preparing, and the mobility lane opens at the completeness gate.

5DecisionSplits at the completeness gate
Evidence complete

Goes to the named mobility lead.

Anything missing

Adds an employment counsel read first.

Mobility review

The trip is held with its duties, its gaps and the date each one falls due.

Clear · Add evidence · Send to counsel review

Cleared — by the mobility lead

  1. 6Travel and expense records updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedDeadline accuracy, declaration currency, lead corrections and what review found

Corrections

Each mobility-lead correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Filing a posting declaration in any country.
    • Deciding that an exemption covers this trip.
    • Approving the trip and the travel it commits.
    • Naming the person a destination requires.

  • Automation boundary

    Agent acts unaided

    • ✓Date each duty from the country, not the diary.
    • ✓Show which trips are undeclared, unheld or already inside a deadline.
    • ✓Draft the declaration and the A1 request for a named person to file.
    • ✓Read the later claim against the trip declared.

    Nothing is filed or cleared except by a named mobility lead, inside the agreed boundaries.

    • Judging whether a visit is a business trip.
    • Telling an inspector what a record means.
    • Accepting an expense claim as substantiated.
  • Changes to the destination rules or the clocks.

Example output

One trip, annotated

This serves a mobility team whose lead has to clear a trip before it departs; below is one trip exactly as the agent leaves it.

Trip record · single traveller

Illustrative example
Trip
Client meeting, one working day
Recorded as
Proposed, not yet declared
Earliest deadline
Eight days before
Evidence of record
Booking record, 3 August 2026
Confidence
Held undeclared
Held for
The named mobility lead

As received

Taken from the booking record and the project brief on file, and it asserts nothing beyond them.

What the record holds

  • Booking record
  • Project brief
  • Prior A1 on file

Why no declaration here

Filing the declaration is a step that the named mobility lead takes.

What the score decides

Below the configured threshold the trip gets a counsel read before the lead sees it.

Action

ClearAdd evidenceSend to counsel review

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every trip

From the booking that starts it

03Evidence

Where the duty lands

Expense audit for the close reads the claim, and a travel management company serves its client; this is the buyer's own programme, where the duty lands before the trip and the proof arrives after.

01Approved path

The proof arrives afterwards

The claim is what proves the trip happened, and the substantiation rules let it arrive long after the declaration deadline it would have evidenced.

02Human review

What was checked, and not found

Checked in the current text: Article 12 sets a ceiling and no floor, the Practical Guide and Decision A2 set no minimum duration, and the business-trip carve-out sits in an April 2026 agreement not yet adopted.

04

Build an evidence trail

The trip, the declaration filed before it and the person named on it stay together.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Travel and booking systems

    • SAP Concur
    • Navan
    • TravelPerk
    • Amadeus
    • Sabre
    • Amex GBT

  • Expense and card data

    • Corporate card feeds
    • claims
    • Receipts, itineraries and mileage records

  • HR and assignment records

    • Workday
    • SuccessFactors
    • SAP HR
    • BambooHR

Agent

Travel and expense audit

  • Reads the trip
  • Dates the duties
  • Holds for the lead

  • Declaration and A1 portals

    • SIPSI
    • ZKO-3
    • LIMOSA
    • A1 requests and national filing records

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six barriers between the model and the departure

Six barriers on one route, the last the highest. What crosses is drawn in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to listing duties when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and destination rules, and note the version each trip was screened under.Track
L4TraceabilityRecord each trip, the duties dated under it, the evidence held and every change to the file.Record
L3Mobility releaseHold the trip for a named mobility lead; the hold governs release, not whether the file is complete.Gate
L2Duty guardrailsTest each duty against the rule text it was drawn from, and refuse a deadline with no rule under it.Restrict
L1Confidence thresholdsRoute a thin trip file to a counsel read before it reaches the mobility lead.Require review
Model coreTrip screened — the destinations, the duties, the deadlines and what is missing
L1 – L2
Test whether a trip may stand
L3
Leaves the filing to a named mobility lead
L4 – L5
Keep the trip and the declaration behind it
L6
Blocks the trip approval when signals degrade

How Nestack evaluates it

Evaluate the whole screening — not only the trip that clears.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the trip an inspector asks about

  1. E1Final-output evaluationDid the trip record the duties and the deadlines it was actually dated on?
  2. E2Step-level evaluationDid the agent read the current rules, the right destinations and the live itinerary?
  3. E3Tool evaluationDid it read and write the correct trip and the correct traveller record?
  4. E4Confidence calibrationDo low-confidence trips actually attract more mobility-lead corrections?
  5. E5Slice evaluationHow does performance change across specific destinations?
  6. E6Business outcomeHow many trips needed a correction before the lead cleared them?

Floor — the record a clearance rests on

Failure modes

Where each failure originates in the agent

Seven failure modes, each set at the stage where it first shows.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • NE-03

      Stale country rule read

      The instrument read is not the one now in force.

    Stage gathers

    The trips, the countries, the dates and the work

  2. 02 · Reasoning

    2 modes

    • NE-04

      Deadline set from the wrong date

      A duty is dated from booking, not the work.

    • NE-06

      Exemption read as covering

      A carve-out not yet law is worked as live.

    Stage proposes

    The duties, the deadlines, the evidence and gaps

  3. 03 · Tool / write

    2 modes

    • NE-02

      Thin trip file passed on

      A trip moves on without the counsel read.

    • NE-05

      Duty bound to wrong trip

      The filing is held against another trip.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • NE-01

      Cleared, duty unrecorded

      The record shows clearance but not the missing declaration.

    Stage returns

    The trip a lead clears and an inspector reads

  5. 05 · Change / Version

    1 mode

    • NE-07

      Silent rule drift

      A national deadline moves while the stored trip keeps the old one.

    Stage tracks

    Model, prompt, duty rules and country records

  • Sev-1 · a trip taken with no duty checked
  • Sev-2 · a declaration falls due unfiled
  • Sev-3 · source degrades, trip held back
See the complete Operations failure-mode catalogue

Affected slices

Short-notice trips absorb the corrections

A destination-level declaration figure can read clean while trips booked inside a week carry most of the corrections. Nestack reports the correction rate by destination, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Trips booked inside a week10.4%3.7×Review
Multi-country itineraries7.4%2.6×Review
Service delivered on site4.6%1.6×Watch
Repeat single-country visits2.3%0.8×Normal

Bar: correction-rate lift vs. repeat-visit baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What a late declaration costs

A cycle closes when the trip declared after arrival is a regression case. That suite is what the next trip cleared is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on trips booked inside a week.

  2. 02

    Diagnose

    The flight booked on Thursday for work starting Monday in a country wanting eight days of notice is worked back until one cause is left standing.

  3. 03

    Improve

    Number the change; the trips that forced it are attached to it.

  4. 04

    Verify

    Each touched trip case is run once more, and one red stops it.

  5. 05

    Learn

    The case is kept for good, and the declaration rules move with it.

  6. Learn → Detect

    The return edge. The next trip screened runs against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, trip screening, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Destination-duty and automation-boundary work.Week 1
  2. 02Booking, expense and HR sources.Week 1, Week 2
  3. 03Country-to-duty and deadline-derivation mapping.Week 1, Week 2
  4. 04Trip and expense ingestion.Week 2
  5. 05Duty, deadline and traveller binding.Week 2, Week 3
  6. 06Completeness scoring and review routing.Week 3
  7. 07Mobility clearance workflow.Week 3, Week 4
  8. 08Travel-system integration.Week 5
  9. 09Declaration and A1 cases.Week 4, Week 5
  10. 10Guardrails and pre-trip controls.Week 4
  11. 11Trip-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne programme, one countryProductionProduction travel workflowAdvancedMultiple entities / countries
Introduced at Pilot
Trip screening to your destinations✓✓✓
Named mobility clearance✓✓✓
Traveller-population baseline✓✓✓
Introduced at Production
Reporting by destination—✓✓
Mobility review workflow in your systems—✓✓
Approved write-back—✓✓
Booking-and-card integration—✓✓
Introduced at Advanced
Multi-country duty sets——✓
Cross-entity trip packs——✓
Large traveller populations——✓
Multi-country declaration controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, trip volume, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your live destinations and the duty each one creates

    Duty capture and deadline derivation

    Week 1

  2. 02Representative trips, claims and declarations

    Evidence binding, deadline logic and the trip baseline

    Week 2

  3. 03Your travel policy and the leads it names

    Duty mapping, deadline derivation and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Booking, expense and HR source assessment, then integration setup

    Week 2

  5. 05Trips you would not want reconstructed

    Declaration cases and the evaluation run

    Week 4

  6. 06What no expense record may cure

    Completeness scoring, review routing, guardrails and release controls

    Week 3

  7. 07A named mobility lead who clears the trip

    Release to the mobility lead, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

The columns are counted, not drawn to fit; a pair in the fifth week means a pair actually runs.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Trip discovery, deadline derivation and the automation boundary

W2Source integration and the traveller-population baseline

W3Declaration binding, completeness logic and release controls

W4Evaluation suite, declaration cases and failure-mode testing

W5Travel-system integration, pilot trips and targeted corrections

W6One travel quarter run under the mobility lead, then Agent Care handover

Reading the band
Each bar covers only the weeks its own work is named for, and week five is shared by design.
At the end of W6
When the declaration record validates, Agent Care takes the agent on.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Operations AI agent

Build a travel and expense audit agent around the duty that lands before the trip.

The duty lands before the trip and the proof arrives weeks after it, and both are lawful. Show us your last quarter of trips and the declarations filed against them. Where a clock started before anyone looked, the agent says so, and a named mobility lead still files and clears.

Typical delivery
4–6 weeks
Deployment
Your trips
Mobility lead
Before the trip
After launch
Agent Care available

Nestack Agents · Travel and expense audit

AGT-OP-13 · Agent Care available after launch

Build a travel and expense agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your trips

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.